Hard-Easy Effect (Difficulty Effect)

Item Difficulty

Lichtenstein and Fischhoff (1977) established what is now called the hard–easy effect, or sometimes simply the difficulty effect, that general-knowledge question overconfidence tends to be stronger when the questions considered by respondents are hard (in the sense that few people answer them correctly) than when the questions are easy. Indeed, when questions are extremely easy, overconfidence gives way to underconfidence. Now, suppose that the items used by Wright, Phillips, and their collaborators were more difficult for their Asian respondents than for their British counterparts, perhaps because of their cultural focus. Then the greater overconfidence exhibited by the Asian respondents could be nothing more than a manifestation of the hard–easy effect. There are indications that the item samples used in the Wright and Phillips studies were indeed more difficult for the Asian than for the British respondents (e.g., the fact that the Asian groups’ calibration curves in Fig. 15.1 were uniformly lower than that for the British group).

Yates and colleagues (1989) deliberately assembled a pool of items that, on the basis of their content, were expected to be equally difficult in the cultures of both their mainland Chinese and American respondents. After the responses were collected, steps were taken to assure that the item pools analyzed really were equally difficult for the two groups. Even under these conditions, the same kinds of overconfidence differences established previously were observed once more: The Chinese respondents exhibited significantly greater overconfidence than did the Americans. Thus, although it remains plausible that item difficulty contributed to the Asian–British differences in overconfidence reported by the Wright and Phillips group, it is unlikely that they could have explained those differences fully.